trigger software validation olga igonkina (u.oregon), ricardo gonçalo (rhul) tapm open meeting –...
TRANSCRIPT
Trigger Software Validation
Olga Igonkina (U.Oregon), Ricardo Gonçalo (RHUL)TAPM Open Meeting – Mar. 1, 2007
Outline:• Status of infrastructure• Schedule for release 13• RDO data for development
and validation
1 Mar. 07 TAPM Open Meeting 2
Automatic tests• Need to address:
– Algorithm validation:• Already quite some work was done (see e.g. Danilo’s talk)
– Integration tests: does it all work together?• Will use the new TrigDecision class (hmm…also under development)
to compare nightly RTT results with reference• Steering monitoring may also be very useful
• Trying to use the Monitoring machinery (under development) for automatic validation tests in RTT: some – Some good developments:
• Tomasz produced IMonitoredAlgo and GenericMonitoringTool to instrument hypo classes
• Meng Wang producing a parallel tool: HLTMonitoringTool (see Data Quality Workshop yesterday)
• looking at joining efforts– Aim to have this working 2 weeks before rel.13 (…i.e. next week)
1 Mar. 07 TAPM Open Meeting 3
Example of histogram for overall validation
1 Mar. 07 TAPM Open Meeting 4
Schedule for release 13
• Machinery would ideally start to be used 2 weeks before release to guide fixes
• An example of hypo instrumented for monitoring will be provided next week and put in RTT
• Developers will be asked to instrument their algorithms fast so that they can be used in RTT (perhaps mostly/only hypos for now)
• By the 12th – 13th March, most hypos should be ready
• This leaves little more than 1 week before rel.13
1 Mar. 07 TAPM Open Meeting 5
RDOs for trigger validation• There are plans to write RDOs to tape immediately after reconstruction
• Only 10% of the HITS for some datasets would be kept on disk
• The trigger cannot run from HITS at present: it requires HITSRDO + RDOESD
• Each slice should compile a list of needed RDO samples so that the TAPM can request that they are kept
• Please consider that lots of background and some signal may be more important than lots of signal and not enough background
• Example from tau slice:– At least 200 kevents from samples J0, J1, J2, J3 – Around 300 kevents of min.bias (5001)– 20 kevents from samples W had (5107), Z (5188), A0 (5862)